Are Stereotypes Leading LLMs' Zero-Shot Stance Detection ?

TL;DR

This paper investigates how large language models exhibit stereotypes in zero-shot stance detection, revealing biases linked to dialect, vernacular, and text complexity that influence their decision-making.

Contribution

It introduces an automatic annotation method to analyze bias in LLMs' stance detection, highlighting specific stereotypes related to dialect and readability.

Findings

LLMs associate low text complexity with pro-marijuana views

African American dialect linked to opposition to Donald Trump in LLM decisions

Significant stereotypes found in zero-shot stance detection by LLMs

Abstract

Large Language Models inherit stereotypes from their pretraining data, leading to biased behavior toward certain social groups in many Natural Language Processing tasks, such as hateful speech detection or sentiment analysis. Surprisingly, the evaluation of this kind of bias in stance detection methods has been largely overlooked by the community. Stance Detection involves labeling a statement as being against, in favor, or neutral towards a specific target and is among the most sensitive NLP tasks, as it often relates to political leanings. In this paper, we focus on the bias of Large Language Models when performing stance detection in a zero-shot setting. We automatically annotate posts in pre-existing stance detection datasets with two attributes: dialect or vernacular of a specific group and text complexity/readability, to investigate whether these attributes influence the model's stance detection decisions. Our results show that LLMs exhibit significant stereotypes in stance detection tasks, such as incorrectly associating pro-marijuana views with low text complexity and African American dialect with opposition to Donald Trump.